Rapid tracking through strongly scattering media with physics-informed neuromorphic speckle analysis

Researchers from the University of Hong Kong, led by Yuqing Cao, have developed computational neuromorphic tracking (CNT), a physics-informed framework that merges asynchronous event sensing with task-driven speckle analysis. Unlike traditional frame-based cameras that sacrifice signal-to-noise ratio for temporal resolution, CNT uses a neuromorphic sensor that only records changes in the scene, dramatically reducing data load and latency. The system formulates speckle patterns—the random interference of scattered light—as a spatiotemporal representation, jointly optimizing temporal and spatial parameters to maximize tracking stability in extreme conditions.

In experiments, CNT achieved robust motion tracking of objects moving 10x faster and under 10x dimmer illumination than conventional systems. This breakthrough expands the operational regime for tracking through strongly scattering media, such as fog, smoke, or biological tissue. Potential applications include autonomous driving in adverse weather, medical imaging through tissue, and surveillance in low-light environments. The work was published on arXiv (2604.25310) and represents a significant step toward efficient, scalable solutions for demanding real-world scenarios involving rapid motion and poor visibility.